BACKGROUND AND AIMS: Endoscopic therapy is the standard treatment for high-grade dysplasia and some cases of T1a esophageal adenocarcinoma (EAC), but it is not appropriate for deeply invasive disease. Data on the value of EUS for patient selection for endoscopic or surgical resection are conflicting. We investigated the outcome of esophageal EUS for the staging and treatment selection of patients with treatment-naive, premalignant Barrett's esophagus (BE) and suspected superficial EAC. METHODS: We retrospectively reviewed consecutive patients who underwent EUS for staging of treatment-naive, suspected premalignant BE and superficial EAC from January 2006 to June 2014. All patients referred for endoscopic therapy routinely underwent EUS. Patients with esophageal masses, squamous cell cancers, previous neoadjuvant therapy, or unrelated pathologies were excluded. Each patient's final diagnosis was verified by EMR, esophagectomy, or forceps biopsy sampling. Test characteristics of EUS were calculated. RESULTS: Three hundred thirty-five patients (mean age, 68 years; 86% male) with BE, a Prague C mean of 2.8 cm, and a Prague M mean of 4.5 cm were staged (pT0, 78% [6% nondysplastic, 24% low-grade dysplasia, 42% high-grade dysplasia]; pT1a, 14%; pT1b, 7%; and pT2, 1%). Sensitivity, specificity, positive predictive value, negative predictive value, and accuracy for patient selection to endoscopic (T1aN0 or less) or surgical therapy with EUS TN staging were 50%, 93%, 40%, 95%, and 90%, respectively. Comparable rates were achieved for patients with nodular BE. Overstaging occurred in 7% of patients, and EUS selected 11% for incorrect treatment modalities compared with pathologic staging. CONCLUSIONS: This study confirms the limited value of EUS suggested in the latest American College of Gastroenterology guidelines for BE management.